Validation

Functional Validation

The functional suites validate event-bus behavior against the local Kind environment:

• MQTT 3.1.1 protocol compliance
• QoS 0, 1, 2 message delivery
• Retained messages and will messages
• High availability and failover
• Federation between layers (CPC <-> CSC)
• Authentication and authorization (OAuth2, mTLS, NKey)
• Topic-based access control

Local Benchmark Tooling

The local environment includes MQTT benchmark tooling for smoke runs and operator-driven benchmark runs. These tests report the observed behavior of the current deployment; they do not define product targets.

Prerequisite: Performance and benchmark targets require MetalLB or an equivalent LoadBalancer from the local stack. On macOS, start docker-mac-net-connect so the host can reach the MetalLB IPs. Linux hosts normally reach the Docker bridge IPs directly.

Full benchmark targets are sensitive to local host capacity. When they report EOFs, readloop stalls, or success-rate misses, capture host CPU and Kubernetes pod metrics with the run; the smoke targets are the portable local validation path.

The MQTT performance suite exercises combinations of QoS level, retention, and deployment topology:

Federation runs are bidirectional: CPC-to-CSC and CSC-to-CPC.

Reported Metrics

The benchmark tooling reports these measurements:

The separate MQTT benchmark suite reports scenario-level measurements:

• Connection rates and peak concurrent connections
• Message throughput (publish and subscribe rates)
• End-to-end latency percentiles (avg, P50, P90, P97, P99)
• Success rates